Differentiating primary progressive aphasias in a brief sample of connected speech

Abstract

Objective: A brief speech expression protocol that can be administered and scored without special training would aid in the differential diagnosis of the 3 principal forms of primary progressive aphasia (PPA): nonfluent/agrammatic PPA, logopenic variant PPA, and semantic variant PPA.

Methods: We used a picture-description task to elicit a short speech sample, and we evaluated impairments in speech-sound production, speech rate, lexical retrieval, and grammaticality. We compared the results with those obtained by a longer, previously validated protocol and further validated performance with multimodal imaging to assess the neuroanatomical basis of the deficits.

Results: We found different patterns of impaired grammar in each PPA variant, and additional language production features were impaired in each: nonfluent/agrammatic PPA was characterized by speech-sound errors; logopenic variant PPA by dysfluencies (false starts and hesitations); and semantic variant PPA by poor retrieval of nouns. Strong correlations were found between this brief speech sample and a lengthier narrative speech sample. A composite measure of grammaticality and other measures of speech production were correlated with distinct regions of gray matter atrophy and reduced white matter fractional anisotropy in each PPA variant.

Conclusions: These findings provide evidence that large-scale networks are required for fluent, grammatical expression; that these networks can be selectively disrupted in PPA syndromes; and that quantitative analysis of a brief speech sample can reveal the corresponding distinct speech characteristics.

Figure 1. Gray matter atrophy and reduced white matter fractional anisotropy in primary progressive aphasia and regressions relating grammaticality to neuroimaging

(A) Nonfluent/agrammatic primary progressive aphasia (naPPA). (B) Logopenic variant primary progressive aphasia (lvPPA). (C) Semantic variant primary progressive aphasia (svPPA). In the left (left hemisphere) and center (right hemisphere) columns, gray matter atrophy is shown in green (q < 0.025, FDR-corrected), and regressions relating grammaticality to atrophy are shown in red (p < 0.05). On the right, reduced white matter fractional anisotropy is shown in orange (q < 0.01, FDR-corrected, except svPPA p < 0.005 uncorrected), and regressions relating grammaticality to fractional anisotropy are shown in pink (p < 0.01). Yellow arrows highlight regressions in naPPA and lvPPA. FDR = false discovery rate.

Figure 2. Regressions relating speech production variables to gray matter atrophy and reduced white matter fractional anisotropy in primary progressive aphasia

(A) Regressions for speech errors per 100 words in nonfluent/agrammatic primary progressive aphasia (naPPA). (B) Regressions for dysfluencies per 100 words in logopenic variant primary progressive aphasia (lvPPA). (C) Regressions for nouns per 100 words in semantic variant primary progressive aphasia (svPPA). In the left (left hemisphere) and center (right hemisphere) columns, gray matter atrophy is shown in green (q < 0.025, FDR-corrected), and regressions are in other colors (p < 0.05). On the right, reduced white matter fractional anisotropy is shown in orange (q < 0.01, FDR-corrected, except svPPA p < 0.005 uncorrected), and regressions are shown in pink (p < 0.01). Right hemisphere gray matter atrophy is not shown in panels B and C and white matter reduced fractional anisotropy is not shown in panel B because there were no significant regression results. FDR = false discovery rate.